Skip to content

Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.

Add parameterization for auto mode in PDF analysis tests and enhance remote parse inference progress test

myhloli54373d6fa863
diff --git a/tests/unittest/test_pdf_analyze_block_content.py b/tests/unittest/test_pdf_analyze_block_content.py
index d9fa9553..b1ac158b 100644
--- a/tests/unittest/test_pdf_analyze_block_content.py
+++ b/tests/unittest/test_pdf_analyze_block_content.py
@@ -129,6 +129,7 @@ def test_convert_vlm_results_deep_copies_projected_mutable_fields() -> None:
assert source_cell_merge == [1, 0]
[email protected]("auto_mode", [False, True])
@pytest.mark.parametrize(
("effort", "parse_mode", "vlm_method_name"),
[
@@ -143,6 +144,7 @@ def test_doc_analyze_converts_vlm_results_before_downstream_processing(
effort: str,
parse_mode: str,
vlm_method_name: str,
+ auto_mode: bool,
) -> None:
"""验证 high/xhigh 的 TXT/OCR 路径都只向后续阶段传递原生 list/dict。"""
source_block = VlmContentBlock(
@@ -154,6 +156,7 @@ def test_doc_analyze_converts_vlm_results_before_downstream_processing(
source_page = ExtractResult([source_block], layout_scored=object())
fake_document = MagicMock()
fake_document.page_count = 1
+ fake_document.classify.return_value = parse_mode
fake_document.__getitem__.return_value = MagicMock(size=(612.0, 792.0))
hybrid_model = MagicMock()
@@ -222,9 +225,11 @@ def test_doc_analyze_converts_vlm_results_before_downstream_processing(
middle_json, model_json = analyze.doc_analyze(
b"fake-pdf",
effort=effort, # type: ignore[arg-type]
- parse_mode=parse_mode, # type: ignore[arg-type]
+ parse_mode="auto" if auto_mode else parse_mode, # type: ignore[arg-type]
)
+ assert fake_document.classify.call_count == int(auto_mode)
+ getattr(vlm_predictor, vlm_method_name).assert_called_once()
assert isinstance(model_json, ModelJson)
assert type(model_json.pages) is list
assert type(model_json.pages[0]) is list
diff --git a/tests/unittest/test_vlm_remote.py b/tests/unittest/test_vlm_remote.py
index c7511b4c..ec5e6344 100644
--- a/tests/unittest/test_vlm_remote.py
+++ b/tests/unittest/test_vlm_remote.py
@@ -373,3 +373,38 @@ def test_low_efforts_never_create_remote_client(
assert result.effort == effort
assert openai_server.requests == []
remote_factory.assert_not_called()
+
+
[email protected]("tier", ["standard", "advanced"])
[email protected]("async_mode", [False, True])
[email protected]("enabled", [False, True])
+def test_remote_parse_inference_progress(
+ openai_server: _OpenAIServer,
+ hybrid_stub: None,
+ tmp_path: Path,
+ monkeypatch: pytest.MonkeyPatch,
+ capsys: pytest.CaptureFixture[str],
+ tier: str,
+ async_mode: bool,
+ enabled: bool,
+) -> None:
+ """真实 PDF 经同步/异步解析和 loopback HTTP 推理后,终端仅出现所属阶段的进度条。"""
+ from functools import partial
+ from mineru_vl_utils import MinerUClient
+ from mineru.model.vlm import runtime
+
+ monkeypatch.setattr(runtime, "MinerUClient", partial(MinerUClient, use_tqdm=enabled))
+ monkeypatch.setattr(config.model, "vlm", _settings(openai_server))
+ source = _pdf_input(tmp_path)
+ result = (
+ asyncio.run(parse_async(source, tier=tier, ocr_mode="ocr")) if async_mode else parse(source, tier=tier, ocr_mode="ocr")
+ )
+ assert "Remote VLM text" in result.markdown()
+ stderr = capsys.readouterr().err
+ description = "VLM Predict" if tier == "standard" else "Two Step Extraction"
+ assert (description in stderr) is enabled
+ if enabled:
+ assert "100%" in stderr
+ if tier == "advanced":
+ assert "VLM Predict" not in stderr
+ assert any(body is not None for _, _, body in openai_server.requests)